Windows: Interview with Tandy Trower
Interview with Mr. Tandy Trower
Tandy Trower joined Microsoft in 1981 and transferred to the Windows development team in January 1985. As the project manager, he wrapped up development and made sure Windows 1.0 would ship in the fall of 1985. He went on to lead the development of Windows 2.0, to be released in 1987, and left the Windows team afterward.
Stay Forever: Can you tell us how you got the assignment to lead the Windows project in early 1985?
Tandy Trower (TT): In late 1984, I was product manager for Microsoft’s programming languages, which at that time not only included Microsoft BASIC, which was pretty much a standard on all PCs, but also COBOL, FORTRAN, Pascal, C, Lisp, and MacroAssembler for 8086/8088. With the exception of BASIC, it was an odd assignment coming from Atari’s Personal Computer Division, where I worked as a product manager, where I managed consumer titles for Atari’s 400 and 800 Personal Computers.
I actually got my start in the PC tech world by learning how to program in Microsoft BASIC, first on a Sol Processor Tech PC, then a Radio Shack TRS-80 Model II, and finally an Apple II. However, Atari had opted for a different vendor for the BASIC on their PCs. That BASIC was designed to run from a cartridge on 16K Atari 400s, and it was not entirely compatible with Microsoft BASIC which ran everywhere else. So I recommended to Atari’s management that we acquire a license to Microsoft BASIC. That resulted in a visit from Bill Gates, who flew down from Seattle to negotiate the features and schedule. When he walked into the room everyone was impressed with his technical savvy despite being so young (about 5 years younger than I was). In fact, he so impressed the Atari execs that they subsequently flew up in the company’s private jet hoping to acquire Microsoft for the growing Warner Communications empire. However, Bill and Paul had already inked a deal with IBM for the PC that was to debut in 1981, so it obviously didn’t happen.
Once the project was signed by both companies, I spent the next 6 months managing the project on the Atari side, then took a brief vacation with my young wife to backpack in the North Cascades and decided that we’d like to move back to the Northwest (as we had lived in the greater Seattle area prior to living in the Bay Area).
Atari had turned into a bit of a mess as they were still struggling on how to beat Apple in the home PC market. I had 3 different bosses in the span of 6 months as Atari kept trying to find the right person. So, I sent a letter (typed at that time on a typewriter) to Bill Gates, asking if Microsoft had any openings they might consider me for. The timing was opportune because Microsoft was just starting to staff up to meet the increased demands coming from the growing PC market. After receiving my offer in August of 1981, I started at the end of September.
At the time, Microsoft was still a relatively small company, with less than 100 employees. Offices were spread across a couple of floors in separate office buildings in downtown Bellevue. I shared an office with the Marketing Communications Director for the first couple of months, before the company moved to a newly constructed building just off of I-520 and conveniently for Gates, adjacent to a Burgermaster restaurant.
My initial assignment included all forms of Microsoft BASIC (since by then I was very familiar with the language), a few educational software products (i.e. Typing Tutor), hardware cards for the Apple II (designed by Paul Allen), and a few games (which would also include Microsoft Decathlon and the first version of Flight Simulator for the IBM PC). And now I was on the other side of the Atari BASIC project managing its successful completion from the Microsoft side.
These assignments changed a few months later when we lost one of our product managers, and I was asked to take on Microsoft COBOL as well (despite that I had no familiarity with the language). By spring of the following year, that had expanded to include the rest of the programming language including the company’s first C compiler, while hardware, games, and educational products moved on to new product managers.
Late in 1984, Turbo Pascal was introduced by Phillipe Kahn for only $50. It sold like hotcakes. It clearly worried Gates that Kahn’s Turbo Pascal might displace BASIC as the default programming language on PCs. It also didn’t help that Kahn’s somewhat arrogant personality irritated Gates as well. So at one of those infamous product reviews where Gates tended to be harsh, his questions turned to me on how I was addressing this. I tried to point out that Microsoft’s Pascal was not intended to be a cheap development tool like Kahn was hawking, as it had been developed to be the primary programming language for Microsoft (this was long before C#). However, this did not mollify Gates in the least. Instead, it only increased his anger, as he demanded to know why I was allowing Kahn to steal the market for Microsoft’s most successful product.
I left the meeting devastated, wondering if I would be fired. So I proactively sent Gates an email confessing that perhaps my assignment to manage programming languages was not my strength. I loved BASIC because it provided a means of allowing a greater audience to access the power of early PCs because it was not as cryptic as other languages, and further my experience at Atari had been on software titles designed for consumers. But the rest of the programming languages clearly appealed to more tech-savvy people. I waited for his response.
Once again, the timing was opportune. Microsoft had announced Windows in 1983, promising to ship it in 1984, and in those days you did not want to be late, or else the tech press would likely label the product as „vaporware“.
But in late 1984, it was clear that the product was clearly not going to ship by year-end. So Gates conferred with Steve Ballmer (my then boss) who oversaw the Systems Group, which not only included Windows, but also the IBM relationship. They offered me a new challenge, the opportunity to transfer over to Windows to get the product out.
However, my first reaction was that this might be a clever attempt to justify my termination. By that time Windows had already gone through 2-3 product managers and had the air of „career death“ about it. To be fair, it was a significant challenge taking a concept of a graphically oriented interface like those developed at Xerox PARC and getting it to run on an IBM PC (4.77 Megahertz, 256K RAM, double floppy drives, and CGA (320×200) display). But Jobs had managed to do that on the diminutive Macintosh. Further, Ballmer wanted another product to license to IBM, as MS-DOS and BASIC had significantly helped boost Microsoft revenues because all other PC clone vendors would also sign up.
While considering the offer, another product manager asked me if I had been offered the position, as he apparently had and passed on it because he didn’t see how Windows would succeed. That only bolstered my assumption that my offer might be a clever plot to terminate me. So I confronted Gates about this. He laughed. Bill reassured me that was not his intention and that he had confidence this would be a better fit and that the company’s future really did depend on this.
So I accepted the position to start in January of 1985. The initial management of Windows was handled by Scott McGregor, a transplant from Xerox PARC, on the dev side, and Leo Nikora from the product management side (Leo was also a former Xerox PARC engineer). But just prior to my transfer, both these individuals left Microsoft. Apparently, Gates and McGregor had several disagreements on the design of Windows. It was McGregor who changed Windows from initially being designed using overlapping windows to „tiled“ windows, using the argument that it optimized screen management when running multiple applications. While McGregor was right with regards to the high-resolution screen displays at Xerox, it was clunky on the far more limited resolution of a CGA screen. One of the first things I proposed was going back to overlapping main applications windows, which didn’t seem like major work as all dialog box windows were already overlapping.
However, Ballmer negated that because the development team told him it would not only add time to the schedule but potentially introduce instability (and possibly because they had already made the change to tiled windows and may have resented having to rework the window management code yet again). As Ballmer wanted Windows to ship no later than June 1985, and the development team had indicated that this might jeopardize that, he overruled that proposed change, offering that I could put it on the top of my list for the next major release.
Ballmer made it very clear that my top objective must be to ship it by that summer. Had Windows been in good shape when I took over, that should have been easy, but I am not certain that Ballmer realized that there were many significant parts not fully implemented. For example, you could not print, there was no font model, no support for localization of text, etc. And as there were no significant applications built for Windows yet, I had added the requirement for a set of applets that would ship with the product. Obviously, despite my reputation and a lot of hard work from the dev team, the work needed to finish it would require a few more months than that.
Finally, I had not been completely empowered in my role. The development team reported to Ballmer, not me. So I often had to use my best skills of persuasion or negotiation with a team that had already been driven hard over the previous two years. On the other hand, I was assisted by peers like Rick Dill, who managed the Windows SDK (Software Developers Kit), and Paul Davis, who led the Windows Developer Evangelism. I was also assisted by Gabe Newell (now CEO of Valve) who spent many endless hours testing Window, often sleeping in his office, and by Joe Rehfeld on marketing, and by Pam Edstrom and Jonathan Lazarus, who as consultants coordinated the eventual release event which included a „Windows Roast“ presided over by Stewart Alsop, then one of the most respected critics of the PC industry.
Stay Forever: According to books like „The Making of Microsoft“ and „Hard Drive“, which cover Microsoft’s rise to power, development of the „Interface Manager“ started in September 1981 – right after DOS was finished. However, other sources claim that Microsoft started this project after Bill Gates saw the VisiOn presentation on Comdex in November 1982, more than a year later. Can you help to clarify this, when was this project really kicked off?
TT: While the origin of Windows pre-dates my personal involvement in the project, I would highly question the suggestion that the VisiOn presentation motivated Gates. I would suggest that one of the biggest influences was more likely the hiring of Charles Simonyi, who was a Xerox PARC transplant and worked on the software for the Xerox Star and possibly its predecessor. So Simonyi would have had extensive experience with bitmapped, mouse-driven PCs and how applications benefited from that.
Further, shortly after Microsoft moved to its Northup Way address in 1982, Bill Gates purchased a Xerox Star. It was installed in a small office space just down the hall from my office. Gates sent out an email message to all employees that we should stop by and try it out. We all still had Zenith video terminals for email communication on our desks (IBM PCs were still not yet available for everybody in the company) because he saw this as the future of where computers were heading, i.e., large resolution, bitmapped displays, that used a mouse for interaction. While I was focused on other products at this point, I did stop by and tinker with it and the applications. At that point though, considering the limitation of displays for the IBM PCs and that mice for PCs were not available yet, it seemed somewhat futuristic. But Microsoft was already moving in that direction with the development of their first computer mouse which debuted in 1983 and had already been working on prototypes on what was to become Windows, though, at that point, it was called „Interface Manager“, only renamed to „Windows“ in 1983.
I can’t recall the history of VisiOn, but at that time Software Arts was most known for VisiCalc, which sold a lot of Apple II computers, but was eventually displaced by Lotus when the IBM PC was debuted in 1981, and WordPerfect owned the PC-based word processing market at that time. If Software Arts demoed VisiOn in November 1982, it doesn’t seem like it could have been more than a prototype. Again, at that time the big app companies were Lotus (1-2-3), MicroPro (WordStar), and Ashton Tate (dBase) with Microsoft somewhat in the second position, working hard to compete with Multiplan and Word. So you’d probably have to ask Gates what most influenced him, but I strongly suspect it was Simonyi or possibly his knowledge of what was happening at PARC.
Early prototypes of Windows may have been initially implemented in Pascal (remember I mentioned that Microsoft’s Pascal compiler was initially groomed to be Microsoft’s core development language). Those prototypes featured overlapping windows on a bitmapped display that could be accessed by a mouse or the keyboard. Dummy apps illustrated what word processing and spreadsheets (Microsoft only had MS-DOS versions of Word and Multiplan at that time) might look like. I can’t recall how functional those prototypes were but post my first experiences with the Xerox Star, it was the first example I recall seeing on a PC, and long before I ever saw a Macintosh (this is important to remember from the Apple lawsuit perspective). By the time I transferred over to the project in 1985, the actual code for Windows had now been implemented in C, as by that time, the Microsoft languages dev team had created this increasingly popular language.
In 1983, Gates also agreed to develop applications for the Macintosh. While there was hardware on-site at that time, it was strictly restricted only to the Application dev team. So I don’t recall seeing a Macintosh until Apple’s Superbowl commercial in January 1984, though I might have seen an Apple Lisa by then, though that product never seemed like it could be successful considering its cost. (Apparently, Jobs thought so too as he pivoted Apple from the Lisa to the Macintosh.)
In any case, Gates‘ vision that GUI would be a core part of the PC evolution was evident to me very early in the 1980s, but I don’t really know what were the key factors that contributed to this.
Stay Forever: What was the main motivation behind the „Interface Manager“ project? Was it the creation of a graphical interface with mouse control, a unified driver interface, establishing program interface and usability standards, creating an entirely new operating system or DOS application multitasking (or, well, something else)?
TT: Again, I was not involved in the original motivation for the product. So, I can’t say precisely what it was, but from what I understood at the time Gates clearly envisioned going beyond command-line interfaces and delivering the benefits of GUI. That said, I suspect many in the industry recognized that character-based screens and keyboard-based input weren’t the ideal user interface. However, it would take a little time for PC hardware to have sufficient graphics capabilities (and memory to drive it).
Again, I suspect that Simonyi’s experience with how GUI transformed the user interaction likely had some role. Gates always very much respected Charles‘ input, especially in his role as the architect for Microsoft’s applications development.
As I noted, for me, the appeal was similar to what BASIC had offered, that is, enabling access to PC technology to a wider range of people, not just those who were comfortable with bits and bytes.
Stay Forever: You said, „Ballmer very much wanted another product to license to IBM“ – so I guess Microsoft didn’t know IBM was already working on TopView (announced in August 1984)? If they had, would Windows have happened?
TT: Prior to my transfer to Windows, my knowledge and involvement with the ongoing relationship with IBM were minimal. However, I do recall that Ballmer heavily lobbied IBM to license Windows. I believe from his perspective this had been an effective strategy; i.e. license to IBM, and you can then license to all those many PC (clone) companies too.
To that end, Steve had a dedicated liaison team to interact with IBM that seemed to be in daily communication with IBM. That team even included some IBM transplants.
So I don’t know when Ballmer/Gates/Microsoft may have become aware of TopView, but at the time I joined the Windows team in 1985, it was clear IBM’s interest in licensing Windows was not very high, and that a character-based solution was preferred. Again, this was partially because most early PCs had minimal (or no) graphics capability and it was clear that even if you equipped a PC with graphics, it would be somewhat of a performance hit to display text as graphics rather than characters just from the perspective of memory.
But not only IBM considered character-based software better for early PC configurations, so did Lotus and WordPerfect. However, IBM did recognize that PC business users would want to view and transfer information between applications, but they didn’t consider it worth the cost of requiring graphics to do that, especially since the most successful business applications didn’t require high-resolution graphics capabilities.
However, even with that, getting DOS applications to load and play nicely together was not easy, as most were designed to maximize the resources of a PC (for example, memory). So Microsoft acquired Nathan Myhrvold’s Dynamical Systems to ensure that the company understood how to do this as well as IBM, and possibly to demonstrate that Microsoft could offer IBM an alternative that could match or beat whatever benefit TopView offered.
So while I can’t say when Gates/Ballmer became aware of TopView, it was clear that it did negate Gates’s objective to develop and ship Windows, as the company initially released the product directly in spite of IBM’s initial disinterest in licensing it. Gates clearly envisioned GUI as being essential for the future and poured resources into it. However, when it became clear IBM would not license the first release of Windows, Ballmer’s strategy then shifted to persuade IBM to jointly develop the successor to MS-DOS, resulting in the Joint Development Agreement on OS/2, which was to include a component called Presentation Manager that would deliver similar capabilities as an integral part of the OS. Having failed to get IBM to license Windows, this was Steve’s new strategy. He even told me that Windows 2.0 would be the end. However, it was clear that Gates was not going to rely on this and required Windows development to continue even with the ongoing development of OS/2.
Stay Forever: Who was the main target audience for Windows 1.0 – OEMs or PC users? Home users or office workers? Microsoft tried to push into the end-user market back then, but I would guess a multi-tasking environment was primarily for office workers.
TT: The obvious target market was end-users, more specifically, any that were the potential market for Microsoft’s core applications (Excel, Word, PowerPoint). You already know that Gates‘ vision included a „computer on every desktop and in every home“. He obviously understood that you had to win the hearts and minds of the office worker first for those applications to find their way into the home. But remember that prior to Apple’s announcement of the Macintosh in 1984, Gates had already committed substantial development resources to developing GUI apps for that platform. PCs themselves were still making that transition from office to home and many early home PCs didn’t really succeed that well because software was typically incompatible from one machine to the other. So while Apple, Commodore, and Atari might have had some initial success, that all changed with both the introductions of the IBM PC and Apple Macintosh.
Stay Forever: A technical question: Windows brought a unified driver interface for displays and printers, ending the necessity to bundle app-specific drivers with every application. But couldn’t this have been achieved with unified DOS TSR drivers as well?
TT: I am not the best to respond to this as I was never involved in the development details, but my impression was that TSR was a hack. Further, while the idea of unified drivers seems an obvious solution for the diversity of hardware, it was not so simple even for Microsoft to achieve this in Windows. At that time, you had everything from daisy-wheel printers to dot-matrix to laser printers coming on the market, and PC display graphics resolutions were still evolving as well. The initial BM CGA and EGA graphics adapters didn’t even have square pixels. So mapping an image from one device to another, each with different resolutions was pretty challenging. Even today I am willing to bet that one of the biggest support challenges for Microsoft is the diversity of displays and drivers that Windows needs to support.
Stay Forever: One source claimed that you, personally, „made it a requirement that Windows could run existing DOS applications.“ Is that correct, wasn’t DOS app compatibility one of the original design goals? How could Windows have survived without that?
TT: I wish I could take credit for making that an objective, but no. In fact, running MS-DOS applications was one of the hardest features to support because, as mentioned, DOS application developers frequently did hacks to maximize memory, which was a carryover from pre-DOS days when applications could assume they were the only software running. Frankly, if I could have cut that requirement, I might have considered it. But it was obvious that for Windows to initially succeed it had to support reasonable compatibility with existing popular apps. Could Windows have survived without it? It would have been hard. Apple obviously did with the Mac, but in large part that was because Microsoft helped by delivering serious business apps from Day 1. But even so, Microsoft was not initially able to do both, and Windows had to wait until version 2.0 before it would have a significant application (Excel) that motivated users to move away from DOS apps.
Stay Forever: You said – as do several other sources – that Scott McGregor removed the „overlapping windows“ feature from Windows 1.0 Beta because he wanted a „tidy“ desktop. Was this decision based on testing feedback or other data or was it just his personal taste? Other sources speculate that a deal with Apple was the main reason for this change (fearing a lawsuit, and rightfully so, as we saw with Windows 2.0), could you elaborate on that?
TT: I think you’d need to ask McGregor to get an accurate reason why. But again, remember he had come from Xerox PARC as well where he saw the benefit of tiled windows, though that was on screen resolutions (processors, memory, and storage) much greater than that of a typical PC. As McGregor left the week before I officially transferred to Windows, we never had an opportunity to discuss this. But I NEVER recall anyone (Gates/Ballmer/others) ever telling me that it must stay tiled because of Apple. The only reason given for why I couldn’t switch it back to overlapping in Windows 1.0 was the potential development and testing impact. Gates/Ballmer agreed that I could make it the highest priority for Windows 2.0. That said, you can see that in later versions of Windows, Microsoft did add support for window management.
Again, I never considered it a part of my objective to make Windows look like the Mac. I might have considered input from our Applications Division on what their requirements were, but my goal was to try to create the best GUI and I saw shortcomings in the Mac GUI. In fact, post Windows 2.0, I continually pushed for enhancements to the UI, composing a detailed specification that I presented to Jim Allchin when he started that attempted to address some of the inconsistencies I saw in Windows and Windows apps. At that point, my role had evolved to where it was my job to push for improved consistency and usability across all Microsoft software, which earned me the title of the „UI cop“.
Stay Forever: When the Windows project was launched, EGA cards were years away and the majority of IBM PCs had the monochrome, text-mode exclusive MDA card (Monochrome Display Adapter) installed. Early Beta screenshots also don’t show icons, just text-based menus. The final release, of course, did use bitmap icons and omitted MDA support. Was MDA support planned at some point or was Windows always envisioned as a graphics-based environment?
TT: I don’t recall seeing any version of Windows (prototype or otherwise) that presented information in text mode. Gates definitely believed that graphics was the future and that it was just a matter of time. In fact, if my memory serves, even early versions of Word (pre-Windows) may have offered the option of running in text or graphics modes (maybe that’s what you may have seen references to), but this was to compete with the fact that neither Lotus nor WordPerfect did not require graphics.
Stay Forever: Both the Atari ST with GEM and the Commodore Amiga with Intuition/Workbench – pretty advanced UIs – were released almost half a year before Windows, did they have any impact on the Windows design?
TT: No, I can’t recall ever looking at either of those during my time on Windows 1.0. As mentioned, I didn’t have much time or resources to consider many visual changes to it. Even getting Gates/Ballmer to agree on the creation of the „desktop applets“ was a significant win, which in many cases meant borrowing dev resources wherever I could. This earned me a bad reputation with the dev manager for the Applications Division when I proposed to acquire his group’s early prototype for Word as an included applet (which I re-christened as Windows Write).
When I did take a look at GEM, it seemed it was more designed to be a Mac UI clone. So I never saw much value in what it offered. I was more impressed with the elegance of the Metaphor UI which was more of a purer derivation of some of the Xerox PARC work as it didn’t have some of the baggage that DOS apps sometimes brought to Windows.
I think that it is hard to imagine now what a difference GUI required in application design. Any attempts to just port a command-line application to Windows failed. Microsoft benefited greatly from learning how to make this transition when it developed its applications for the Macintosh, which also benefited Apple.
Stay Forever: Why was Windows 1.0 so late before you took over, what was the main reason for the delays?
TT: I think it was a very ambitious project that tried to encompass many goals. As you point out in one of your earlier questions, there were so many goals and frankly, the hardware (processor, screen res, memory, disk storage) really was not there yet. Ballmer had made it a goal that Windows 1.0 had to run on a PC with CGA, dual floppies, and 256K RAM because that was still considered a high-end PC at the time, but frankly, it didn’t run that well on that (minimal memory Macs weren’t much better).
The good news is that PC technology in terms of processor speed, graphics capability, and disk storage was rapidly improving. Early competition forced companies to find ways to differentiate their hardware while maintaining compatibility with existing software.
In any case, as I have learned with every project I have ever undertaken, even with the best, clearest objectives, it is impossible to anticipate all the issues you may encounter. For example, in May of 1985, with the Ballmer June deadline rapidly approaching, the developer in charge of the memory manager inside Windows (key to all its operations) told me he had discovered a significant flaw in its design and would have to rewrite it. I told him he could not do that at this point, but he made a convincing argument that Windows could not survive without it. So I agreed despite that it would invalidate most of the testing we had done up to that point.
June came and we were obviously not done. Ballmer was not pleased. Gates was not pleased. So I came up with a revised plan that we would create a „Premiere Release“ for the press and reviewers by August (consider it a kind of late beta) and we retargeted for fall. With that, Ballmer announced the reset date of fall („after the leaves fall, but before the snow does“) and contracted with Jonathan Lazarus to come up with a plan that might further soothe the trade press that it would really happen this time. All this bought us a bit more time.
Meanwhile, I still had to deal with the fact that outside the „desktop apps“, there were NO significant applications poised to ship on day one, not even from Microsoft’s Apps division. Excel would take another year and a new version of Windows.
Stay Forever: What do you think were the main reasons for the weak third-party software support in Windows 1 and 2?
TT: There were several things. First, Windows consumed resources that diminished what was available for applications. Remember the maximum memory capacity in those days was only 640 Kilobytes. While that was a significant jump over what 8-bit processors could address, DOS application vendors still wanted to use all available memory for their applications‘ code and data.
Second, a GUI requires a redesign. You can’t just port a command-line DOS app and succeed. You need to rethink how you lay out the interface for the functions. It is a bit like sketching with a pencil and then being presented with a set of brushes and paints. So the entire app industry required re-educating themselves on how to appropriately leverage GUI. (This happened with GUI mobile phones.) As many application vendors preferred the larger PC market, they often chose not to invest in developing for the Apple Macintosh.
Third, if you assume that the primary categories were already taken by DOS vendors, it required some creativity to think about what new kind of application could forge new territory. Slideware like PowerPoint might seem like a no-brainer today, but when it was initially developed it created a new category of business application. PageMaker was another. We are spoiled by the diversity that’s in the market today, but back in the mid-80s the software industry was still trying to define itself beyond the basics.
Stay Forever: Why was Reversi included with Windows 1 – was it just supposed to train users on the mouse, or did you expect Windows to become a platform for gaming? Did you approach game developers for support (or offer the Windows runtime for bootloader games)?
TT: Reversi was initially just designed as a test app as well as part of the mini-app suite that offered value. Even early on, most PCs you bought came with some software that enabled the user to do something with it without having to necessarily go out to buy something.
That said, I did imagine that Windows could support any category of software. Games would be a challenge though because at that time developers of very advanced games wanted to talk directly to the graphics memory to squeeze the max performance and Windows didn’t initially have libraries that would enable this. DirectX better opened the door for this.
Stay Forever: Windows 1.0 shipped with a selection of small applications, including Write and Paint, how did those come about?
When I transferred over to the project, there were no real Windows applications at that time beyond a preliminary start on a Paint application (like Mac Paint), a simple text editor (that I renamed Notepad), and a few demo apps. That’s why I quickly defined specs for a set of mini-apps that could match what Apple offered on the Mac as well as what was offered on many DOS TSR plug-in apps. However, these still could not jeopardize the deadline to ship by June.
Stay Forever: Did you expect Write, Paint, or Calculator to be stop-gap solutions until there would be sufficient third-party software support, or did you expect these tools to be kept and updated for decades to come?
TT: Again, I always felt that out-of-the-box you needed something to try out the new platform. Even Mac and iPhone demonstrated this. So I thought/think of them being essential, but also that they would be replaced by more significant apps. Windows Write was no Word for Windows, Windows Paint no Photoshop, and Calculator no replacement for Excel. But they were a great starter set. I am impressed that even today Notepad, Paint, and even Terminal are considered essential applications for Windows (obviously with improvements).
Stay Forever: According to an InfoWorld article from 1986, at a developers‘ conference „Microsoft implied that the technical guts of Windows will eventually be part of DOS“, hinting that this might happen in 1987. Did you actually plan to fuse DOS with Windows at that time (something that didn’t really happen until 1995)?
TT: I knew that the future of Windows was to unify it with MS-DOS, but I also knew that this would take time. As OS/2 was to be initially defined to be that, I knew that it would be beyond the scope of my time managing the product.
Stay Forever: What were the main design goals for Windows 2.0?
TT: After Windows 1.0 was released in November 1985 at COMDEX, at a „roast“ conducted by invited host and then noteworthy tech critic, Stewart Alsop, I looked forward to Ballmer’s promise to address the issues I was not permitted to address (like returning Windows to overlapping windows) because of the objective to ship Windows in 1985.
One of the things I really wanted to fix in Windows 1.0 is that the system font (used to populate title bars, menus, and controls, e.g., buttons) could only be fixed pitch, meaning like CRT text every character was the same width, which made things look like typewriter text. While Windows 1.0 actually did support „proportional fonts“ (where characters vary by font design) the development team indicated it would be a big hit to schedule. Like returning to overlapping windows, this was at the top of my list for Windows 2.0 and did happen. But it accounted for why Windows 1.0 looked somewhat „clunky“. Mac shipped with a propositional system font on day 1 often resulting in mocking comparisons from loyal Mac users.
Stay Forever: Windows 2.0 was developed alongside the joined OS/2 project with IBM. How did this affect the Windows development team? According to several sources, Steve Ballmer wanted to drop Windows development in favor of OS/2 and Presentation Manager.
TT: Remember that Ballmer was not only my boss and the exec in charge of the Systems division at Microsoft (and a close personal friend of Gates), but also Microsoft’s primary account executive for the company’s ongoing relationship with IBM, and Ballmer very much wanted to continue the structure of the business relationship of where MS-DOS and BASIC were installed by default on every IBM PC. While IBM got a pretty sweet deal on their license on these two Microsoft products (as well as Microsoft’s entire suite of products), the number of PC clones who all wanted to replicate IBM’s offering (e.g., Compaq, HP, etc.) generated very significant revenues for the company. But IBM did not receive any share of that revenue, and further execs at IBM soured on their dependency on Microsoft. So while Ballmer wanted to get IBM to license Windows, IBM was not interested. They did not recognize the value of Windows. Its GUI required resources that DOS software did not and they preferred to continue that route with TopView to enable users to run multiple applications in a windowed environment without going full graphical.
But Ballmer was shrewd. He changed the conversation with IBM to focus on what would come next, proposing instead a jointly developed new graphical OS unified with the file system where the companies would share revenue. Ever the persistent salesman, Steve was successful, and the companies signed a JDA in 1985. However, he also wanted to ensure that Microsoft could satisfy IBM’s TopView implementation and Microsoft acquired Nathan Myhrvold’s Dynamical Systems who had created one of many TopView clones. Myhrvold, who had been at UC Berkeley, had a small and very talented team of engineers that complemented Microsoft’s own team, and effectively Myhrvold became Ballmer’s dev manager for Microsoft’s side of the OS/2 project.
However, Ballmer also reassigned most of what had been the Windows 1.0 development team also to the OS/2 project to be combined with Myhrvold’s engineers. I ended up with only a few of the original Windows team members to develop 2.0. However, as Gates wanted the chief objective of Windows 2.0 to be the premiere engine for a PC version of Excel (Excel for Mac had been a huge success), I was given a new development manager from that division and a few additional new engineers to fill the gap. So essentially, we were somewhat almost starting over with Windows 2.0. While it was the same starting codebase, having a new engineering team on the software made the work a bit more challenging and I had to balance my objectives between fixing all those things I had wanted to address in Windows 1.0, Gates‘ requirement for Windows 2.0 to be a great engine for Excel, and one further thing… Gates wanted there to be a consistent UI and transition between Windows and OS/2.
So now I was also required to be Microsoft’s UI liaison on the OS/2 project. This required that I meet regularly with several IBM teams, all of who claimed responsibility for OS/2 UI but didn’t necessarily agree on their end. Further, IBM’s UI people had their own strong opinions on the UI for OS/2. An early requirement was that all menus include the text „Esc = Cancel“ so that users would know how to dismiss them. And it didn’t help to argue against that just based on design opinion. IBM demanded that for any design disagreements Microsoft would need to provide usability evidence. Further, at that point, most of the design details in Windows were based either on Gates‘ opinion, those of our Applications Division, or my own personal experience – none of which IBM considered to be sufficient. So I co-opted some designers and usability people from the group that designed print documentation and help files for all Microsoft products.
This all made the development of Windows 2.0 (which was to ship two years later) as big a challenge as Windows 1.0 had been. Once again, I ended up with a „lost“ summer where my wife and kids saw little of me during this time as I spent the majority of each day (including weekends) with the rest of my team.
But despite the challenges, the project was more predictable, and we hit our target ship date. Everyone got what they wanted, and the first release of Excel for PCs bundled Windows in its installation. At that point, I had a discussion with Ballmer about what comes next. Ballmer told me that I should consider looking to other projects as OS/2 would replace Windows 2.0, so there would be no further need as I was a minor player in the OS/2 project. But Gates felt very differently. Gates was not convinced that the OS/2 project was going to succeed. He had many meetings himself with IBM execs regarding the project, especially related to the fact that IBM didn’t consider the revenue split appropriate as they could provide development logs that showed that IBM engineers generated more lines of code than Microsoft’s engineers. Discussions like this drove Gates crazy as he made the point that code size was not an appropriate measure of contribution or quality. So Gates ordered Windows development to continue but also to ensure that so long as the JDA continued that the UI between the two would be compatible.
Fortunately, for Windows 3.0 the goal was simply to take better advantage of improved processors, memory, and storage for PCs. Little needed to be done on the UI side. So development and management were passed over to the MS-DOS team. To address Gates‘ other goal, I proposed a new role where I would continue as the UI liaison but would found a new organization specifically to focus on UI design and usability. Gates loved that proposal because he could see how that would benefit the company’s other software as well. Concurrent with this, was Microsoft’s hiring of a former IBM executive by the name of Mike Maples. Mike’s hiring had a number of benefits. First, Mike understood the IBM culture from working there and while he wasn’t given responsibility for managing that relationship (he was assigned as the Apps Division head) his connections, not to mention experience with product development helped provide a bit more structure to Microsoft’s product development processes. I ended up transferring over to work for Mike because as the OS/2 project continued to deteriorate, my new group’s focus was to handle the increasing demand from Microsoft’s other groups for the services my group provided. Almost every product wanted access to design and usability resources.
Of course, the OS/2 joint development finally failed, though IBM did release something, Gates‘ insurance on the development of Windows 3.0 in parallel saved that from being a disaster for Microsoft. By then also PC vendors were less concerned about being pure clones and ultimately Ballmer gave up on trying to drive the company’s revenues based on that relationship. Instead, he shifted to the Microsoft Sales organization where he spent the coming years focused on that before ultimately taking over as CEO from Gates.
Stay Forever: Were you involved with the Apple/Microsoft lawsuit about the „look and feel“ of Windows 2.0, which Apple claimed to have copied the Macintosh interface?
Yes, I ended up being focused on this lawsuit that Apple brought against Microsoft and spent about a year with Microsoft’s legal team on its defense. For me, the Apple lawsuit was one of the most painful parts of being associated with Windows, as I felt personally accused of just being a copycat, which was never my goal. I chalk up being deposed by Apple’s legal team as probably even the worst part of my entire career in tech and came to understand that lawsuits can sometimes be less a quest for truth and that this one seemed more about finding specific wording that could be considered implicative about potentially violating Apple’s „look and feel“ copyrights. I was often asked the same question many times in different variants, in what seemed an attempt to get me to say things in a particular way. IMHO it was a total waste of time and money that benefited no one.
Stay Forever: According to some sources, the development of Windows 3.0 started as a side project of only two programmers (David Weise and Murray Sargent), who developed „real“ protected mode multitasking for native Windows applications in 1988. Why wasn’t Windows 3.0 a larger project after the launch of 2.0, did the entire company – including Gates – focus on OS/2 at that time?
TT: Weise was/is a brilliant engineer and as Windows would run much better if it could get beyond the 640K RAM limitation, he was very creative in coming up with alternatives. I, unfortunately, don’t recall Murray’s contribution to this. Weise’s continued effort to push this in Windows 3.0 didn’t require a large dev effort and recall that the majority of the former Windows team were now assigned to help jointly build OS/2. Again, initially, Ballmer’s success in negotiating the JDA with IBM was considered the key to Microsoft’s OS future. However, as this never proceeded successfully, I think Gates wanted to hedge his bets and devote some resources to keep Windows going. That proved to be the right bet.
Stay Forever: Besides „real“ multitasking, what were the main design goals for Windows 3?
TT: More memory, leveraging that Windows could now rely on hard disks being an included PC component. But as I had moved on, you’d have to ask Weise, Phil Barrett, or Adrian King for more insight on the goals. All I recall is that this was intended primarily to be a plumbing improvement and not address any UI changes.
Stay Forever: Looking back, are you proud of what you achieved with Windows 1.0?
TT: When I transferred over to the project, my given objective was to ensure it shipped that year. While we missed Steve’s objective by June 1985, that it did ship that year is a statement to the great team that made it happen.
It had been one of the greatest challenges I faced at that time, and I still count it as one of the greatest challenges of my 40+ year career in PC tech. It didn’t help that the goals for the product exceeded what the typical PC configuration could handle, or that it was at the center of what would result in the eventual ending of the Microsoft-IBM collaborative development effort.
IMHO, Windows is a statement to Microsoft’s tenacity and perseverance to push forward regardless of the issues and obstacles encountered to make Windows the most popular OS on the market today. My part in all this was just to help get those first two releases out the door. I was followed by many other great people like Brad Silverberg who came after and further transformed it into what it is today.
Likewise, it also relied on Gates‘ unwavering commitment to make Windows the foundation of Microsoft’s PC (DOS) platform for its own applications that would eventually dethrone the DOS app kings. In short, Gates bet heavily on Windows being the future of PCs and succeeded even without IBM’s initial endorsement, Lotus, and WordPerfect’s adoption, and in spite of Apple’s lawsuit. Is there any software that started out with such challenges that had as significant an impact on the PC world? And did Gates do all this because he wanted to become a billionaire? Definitely not. Bill truly believed that Windows was a key piece to making PCs and their benefits accessible to everyone in their office and in their homes. If he had not, I would never have accepted the offer to join the project.
I think Windows has been one of the most significant pieces of software that Microsoft developed for the PC community. I’d put BASIC and Office and maybe Azure into that category as well, but without Windows I don’t think the latter two could have had as big an impact. As noted it took a lot of work from a lot of great people not to mention Gates‘ vision to push it forward even when it had shaky beginnings, like a parent patient enough to let their child develop into their potential.
Stay Forever: Thank you so much! Is there anything else you would like to add?
TT: I noticed in your series you interviewed David Fox [co-developer of Zak McKracken and other LucasFilm Games titles], so I thought I might share a connection I have with him.
In college, I majored in psychology as I was interested in human behavior, but out of curiosity had room for a class in FORTRAN. However, the instructor was so bad that it is a wonder that I ever ended up in computer tech. In those days, you created programs using punch cards, so you often had to wait for a free keypunch machine, then submit your deck to the computer center to process your program, typically resulting in a listing that had errors in it so that you had to return to fix your deck (and don’t drop it and get the cards out of order), hoping also that the university’s IBM was not giving priority to grades or class registrations so you could get your results back in time, potentially having to do that cycle another time or two. It didn’t help that my prof spent only the first day in class and his TA merely assigned chapters, barely explaining anything. I only survived because a neighbor, who was a CS major, tutored me through the process. I passed the course but was convinced at the time that computers would never be a part of my future.
But I remained fascinated by new technology and frequented the local library where I happened to see that same issue of Popular Electronics (January 1975) that apparently Paul Allen had read also and took with him to visit Gates at Harvard to convince him that the PC revolution was happening. And if you know that story, that led to Gates leaving Harvard to start Microsoft with Allen and deliver BASIC for the MITS Altair.
While I didn’t have the positive experience that Gates and Allen had, the article and prospect for a „personal“ computer that didn’t require punch cards or a computing center had some appeal, but still as a student, I could not imagine affording the purchase of an Altair (let alone a place to put in in the small studio apartment my wife and I had). But I was in luck. Somehow I learned about this couple, David and Annie Fox, who set up the Marin Computer Center in an unused school building, and offered access to computers like the Altair, Processor Technology Sun, Apple 1, and Commodore PET where you could rent time on the computer for a small fee. I tried it and I was hooked. Admittedly, my first attraction was the games. My dad had purchased the very successful Atari VCS that I had used, but this was different because you could program your own games. Over the coming months, I frequently visited Fox’s computer center and consumed every issue of Byte and Creative Computing.
When I graduated my father offered to buy me a PC as a gift. At the time the best value seemed to be the Radio Shack TRS-80 Model II (16K RAM, cassette data storage). However, within a week, it developed keybounce problems (a common early issue), so I returned it for repair. However, in the meantime, I educated myself on the newly released Apple II and was dazzled with its ability to do color and bitmapped graphics. So I requested a refund on my Model II (which Radio Shack was happy to accommodate), stole a bit from our household budget, and purchased an Apple II. Now I didn’t have to visit David and Annie’s computer center, but I would still often stop by just to get the latest on what was happening. I also ended up joining the local Apple club in San Francisco, where I would occasionally give Andy Hertzfeld (noteworthy Apple developer) a ride back to the UC Berkeley campus in exchange for some of the wizard-y software he had created.
One day while in the Fox computer center, my attention was drawn to their bulletin board where a note was posted by a company looking for Apple II programmers. I applied and got the job, even though I could only program Apple IIs in BASIC. And that was the launch point for my career in the PC world. This job led to the next in Atari’s home computer group, which in turn led to meeting Bill Gates in 1980, and subsequently joining Microsoft in 1981.